That seems counterintuitive to me in the context of modern AI approaches.

How so? Elaborate?

I’m wondering if you could elaborate on that a bit more.

This seems sufficiently explanatory to me, especially the italicized part...

AI can’t come up with enough believable unique names for all the posts they want their AI bots to make

Unbelievable usernames becomes an easy identifier/tag for identifying bot post.

Edit: since this comment got downvoted (as the assumed reply) I thought I would elaborate a bit more.

Basically, we name our user accounts to fit the society we live in's norms, it's naming conventions.

If you just run a bunch of vowels and consonants together, that does not make a username, at least not one that people will recognize as a valid one created by a human being.

Part of how bots are effective is in the quantity of bots that are used. Since it's near zero cost to spin up a new bot to make posts/comments, many can be made.

However people can track the validity of a user name as being a bot versus human by the quantity of the posts/comments the username makes (only so many hours in a day, and human beings are busy with other things besides just posting on Lemmy), so no one single bot can make too many posts/comments at one time.

Because of this, you need a large quantity of unique names, one for each of your bots, and they have to be believable ones by humans, so they're not identified as bots.